analiza aiqbench
VeloBins advances aerial inertial odometry by reframing velocity estimation as a classification task over discretized bins, which simplifies uncertainty modeling and improves robustness during aggressive maneuvers. This approach eliminates the need for separate uncertainty decoders while achieving significant error reductions across diverse flight regimes, including high-speed drone racing. The method demonstrates that explicit error-conditioned supervision can enhance filter consistency without directly optimizing negative log-likelihood losses.
VeloBins: Learning Velocity and Its Uncertainty via Bins and Error-Conditioned Gaussian Labels for Aerial Inertial Odometry
arXiv:2608.29720v1 Announce Type: new Abstract: Inertial odometry (IO) is critical for aerial robots, where aggressive maneuvers and poor lighting degrade visual sensors. Recent learning-based IO methods improve traditional integration-based approaches by learning motion priors from IMU and platform-specific sensors, then fusing the predictions within an extended Kalman filter. However, learning velocity through regression is difficult, while jointly estimating uncertainty with a separate decoder and negative log-likelihood (NLL) loss further complicates training and can lead to over-confident estimates. We introduce VeloBins, which reformulates velocity regression as classification over discretized velocity bins. We decode both the velocity from the bin distribution's expectation and the uncertainty from its variance, removing the need for a separate uncertainty decoder. We further supervise the uncertainty explicitly using an error-conditioned Gaussian label centered at the ground-truth velocity, with a standard deviation set to the velocity error. We evaluate VeloBins on four aerial datasets, ranging from free-form aggressive flights and a 27 g nano-quadrotor to drone racing at over 21~m/s. VeloBins achieves the lowest average errors on all four datasets, reducing velocity, relative trajectory, and absolute trajectory errors by 3-27%, 8-40%, and 6-53%, respectively, compared with the strongest baseline. Notably, the proposed supervision achieves the lowest NLL and best filter consistency despite never optimizing an NLL loss. The code will be available upon acceptance.
Abstract:Inertial odometry (IO) is critical for aerial robots, where aggressive maneuvers and poor lighting degrade visual sensors. Recent learning-based IO methods improve traditional integration-based approaches by learning motion priors from IMU and platform-specific sensors, then fusing the predictions within an extended Kalman filter. However, learning velocity through regression is difficult, while jointly estimating uncertainty with a separate decoder and negative log-likelihood (NLL) loss further complicates training and can lead to over-confident estimates. We introduce VeloBins, which reformulates velocity regression as classification over discretized velocity bins. We decode both the velocity from the bin distribution's expectation and the uncertainty from its variance, removing the need for a separate uncertainty decoder. We further supervise the uncertainty explicitly using an error-conditioned Gaussian label centered at the ground-truth velocity, with a standard deviation set to the velocity error. We evaluate VeloBins on four aerial datasets, ranging from free-form aggressive flights and a 27 g nano-quadrotor to drone racing at over 21~m/s. VeloBins achieves the lowest average errors on all four datasets, reducing velocity, relative trajectory, and absolute trajectory errors by 3-27%, 8-40%, and 6-53%, respectively, compared with the strongest baseline. Notably, the proposed supervision achieves the lowest NLL and best filter consistency despite never optimizing an NLL loss. The code will be available upon acceptance.
8 pages, 6 figures, 4 tables. Supplementary video: this https URL
Robotics (cs.RO)
arXiv:2608.29720 [cs.RO]
(or arXiv:2608.29720v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.29720 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 315 cuvinte
Subiecte conexe
Self-attention summary networks for subsurface velocity-model building from common-image gathers
arXiv · 8 oct. 2026
How to Evaluate General-Purpose Robot Policies for Real-World Deployment
NVIDIA · 12 iul. 2026
Inevitable Uncertainty in Probabilistic World Models
LessWrong · 27 iul. 2026
Generalized Gaussian Temporal Difference Error for Uncertainty-aware Reinforcement Learning
arXiv · 28 iul. 2026